Uploaded image for project: 'Blazegraph (by SYSTAP)'
  1. Blazegraph (by SYSTAP)
  2. BLZG-1159

startHAServices no longer respects environment variables

    Details

      Description

      The 1.5.0 release has introduced unconditional sets of various environment variables in src/resources/bin/config.sh. This is invoked by startHAServices in the same directory. Thus environment variables can no longer be set externally (before invoking startHAServices) to override the defaults.

      This also invalidates the example at http://wiki.bigdata.com/wiki/index.php/HAJournalServer#Deployment since it is setting defaults that in fact have no effect without editing the config.sh file.

      The directions at http://wiki.bigdata.com/wiki/index.php/HAJournalServer#Deployment can not be executed as given. They suggest that the ant "deploy-target" should be used and the tarball unpacked into /var/lib. However, they do not call sufficient attention to the version information in the tarball name and they specify dist/bigdata/bin/startHAServices when the unpacked path will probably be bigdata/bin/startHAServices.

      The checklist for the release (and ideally CI itself) should include validation of the HA3 deployment.

      Workaround: directly edit src/resources/bin/config.sh

        Activity

        Hide
        bradbebee bradbebee added a comment -

        Changed committed to master in SF and private Git. Stage for 1.5.1 release.

        Show
        bradbebee bradbebee added a comment - Changed committed to master in SF and private Git. Stage for 1.5.1 release.
        Hide
        bryanthompson bryanthompson added a comment -

        I am still seeing problems with this. When I setup the environment as follows:

        export wd=/root/git/bigdata
        cd $wd
        #git pull
        ant clean stage
        export FEDNAME=benchmark2
        export LOGICAL_SERVICE_ID=HAJournal-1
        export FED_DIR=$wd
        export LOCATORS="jini://bigdata12/,jini://bigdata13/,jini://bigdata14/"
        # Note: Sharing zookeeper with bigdata15,16,17 cluster.
        export ZK_SERVERS="bigdata15:2081,bigdata16:2081,bigdata17:2081"
        export JETTY_XML=${wd}/dist/bigdata/var/jetty/WEB-INF/jetty.xml
        export JETTY_WEB_XML=${wd}/dist/bigdata/var/jetty/WEB-INF/web.xml
        export JETTY_PORT=8090
        export REPLICATION_FACTOR=3
        export groupCommit=true
        # edit dist/var/config/jini/HAJournal.config before starting. E.g., for BSBM optimizations.
        dist/bigdata/bin/startHAServices
        tail -f HAJournalServer.log | egrep '(ERROR|WARN|FATAL)'
        

        The console reports

        Caused by: java.lang.RuntimeException: java.lang.RuntimeException: Could not connect to zookeeper: state=CONNECTING, configZookeeperClientConfig{ zroot=/my-cluster-1, sessionTimeout=60000, servers=localhost:2181, acl=[31,s{'world,'anyone}
        ]}, elapsed=11012ms
        	at com.bigdata.journal.jini.ha.HAClient$HAConnection.start(HAClient.java:735)
        	at com.bigdata.journal.jini.ha.HAClient$HAConnection.access$100(HAClient.java:462)
        	at com.bigdata.journal.jini.ha.HAClient.connect(HAClient.java:267)
        	at com.bigdata.journal.jini.ha.HAJournalServer$HAQuorumService.getZooKeeper(HAJournalServer.java:4552)
        	at com.bigdata.quorum.zk.ZKQuorumImpl.getZookeeper(ZKQuorumImpl.java:133)
        	at com.bigdata.quorum.zk.ZKQuorumImpl.getLastValidTokenFromQuorumState(ZKQuorumImpl.java:2258)
        	at com.bigdata.quorum.zk.ZKQuorumImpl.getLastValidTokenFromQuorumState(ZKQuorumImpl.java:105)
        	at com.bigdata.quorum.AbstractQuorum.start(AbstractQuorum.java:505)
        	at com.bigdata.journal.jini.ha.HAJournalServer.startUpHook(HAJournalServer.java:942)
        	at com.bigdata.journal.jini.ha.AbstractServer.run(AbstractServer.java:1881)
        	... 9 more
        Caused by: java.lang.RuntimeException: Could not connect to zookeeper: state=CONNECTING, configZookeeperClientConfig{ zroot=/my-cluster-1, sessionTimeout=60000, servers=localhost:2181, acl=[31,s{'world,'anyone}
        ]}, elapsed=11012ms
        	at com.bigdata.journal.jini.ha.HAClient$HAConnection.start(HAClient.java:700)
        	... 18 more
        

        That is, it is still looking at the wrong federation name for the zookeeper root path (zroot).

        Show
        bryanthompson bryanthompson added a comment - I am still seeing problems with this. When I setup the environment as follows: export wd=/root/git/bigdata cd $wd #git pull ant clean stage export FEDNAME=benchmark2 export LOGICAL_SERVICE_ID=HAJournal-1 export FED_DIR=$wd export LOCATORS="jini://bigdata12/,jini://bigdata13/,jini://bigdata14/" # Note: Sharing zookeeper with bigdata15,16,17 cluster. export ZK_SERVERS="bigdata15:2081,bigdata16:2081,bigdata17:2081" export JETTY_XML=${wd}/dist/bigdata/var/jetty/WEB-INF/jetty.xml export JETTY_WEB_XML=${wd}/dist/bigdata/var/jetty/WEB-INF/web.xml export JETTY_PORT=8090 export REPLICATION_FACTOR=3 export groupCommit=true # edit dist/var/config/jini/HAJournal.config before starting. E.g., for BSBM optimizations. dist/bigdata/bin/startHAServices tail -f HAJournalServer.log | egrep '(ERROR|WARN|FATAL)' The console reports Caused by: java.lang.RuntimeException: java.lang.RuntimeException: Could not connect to zookeeper: state=CONNECTING, configZookeeperClientConfig{ zroot=/my-cluster-1, sessionTimeout=60000, servers=localhost:2181, acl=[31,s{'world,'anyone} ]}, elapsed=11012ms at com.bigdata.journal.jini.ha.HAClient$HAConnection.start(HAClient.java:735) at com.bigdata.journal.jini.ha.HAClient$HAConnection.access$100(HAClient.java:462) at com.bigdata.journal.jini.ha.HAClient.connect(HAClient.java:267) at com.bigdata.journal.jini.ha.HAJournalServer$HAQuorumService.getZooKeeper(HAJournalServer.java:4552) at com.bigdata.quorum.zk.ZKQuorumImpl.getZookeeper(ZKQuorumImpl.java:133) at com.bigdata.quorum.zk.ZKQuorumImpl.getLastValidTokenFromQuorumState(ZKQuorumImpl.java:2258) at com.bigdata.quorum.zk.ZKQuorumImpl.getLastValidTokenFromQuorumState(ZKQuorumImpl.java:105) at com.bigdata.quorum.AbstractQuorum.start(AbstractQuorum.java:505) at com.bigdata.journal.jini.ha.HAJournalServer.startUpHook(HAJournalServer.java:942) at com.bigdata.journal.jini.ha.AbstractServer.run(AbstractServer.java:1881) ... 9 more Caused by: java.lang.RuntimeException: Could not connect to zookeeper: state=CONNECTING, configZookeeperClientConfig{ zroot=/my-cluster-1, sessionTimeout=60000, servers=localhost:2181, acl=[31,s{'world,'anyone} ]}, elapsed=11012ms at com.bigdata.journal.jini.ha.HAClient$HAConnection.start(HAClient.java:700) ... 18 more That is, it is still looking at the wrong federation name for the zookeeper root path (zroot).
        Hide
        bryanthompson bryanthompson added a comment -

        Ah. I think that the issue is that the BIGDATA_1_5_0_Ticket_1336 branch was not made from master but from the 1.5.0 release.

        Show
        bryanthompson bryanthompson added a comment - Ah. I think that the issue is that the BIGDATA_1_5_0_Ticket_1336 branch was not made from master but from the 1.5.0 release.

          People

          • Assignee:
            beebs Brad Bebee
            Reporter:
            bryanthompson bryanthompson
          • Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved: